Building intelligent systems from zero to scale.

Agentic AI and semantic data layer for power-market infrastructure.

Designed an IEC CIM-aligned power-system ontology and built agentic AI services surfacing warehouse data to AI agents via MCP and REST/SPARQL — enabling safe LLM access to production power-market data.

• Led design of a power-system ontology (OWL/SHACL, YAML-LinkML) over a Databricks lakehouse, with a stable graph Reader API across pluggable backends.
• Shipped dual-transport service surfacing warehouse data + semantics to AI agents (MCP) and applications (REST/SPARQL), with CI-enforced parity.
• Built SCED scenario engine orchestration — Dagster → Argo → Metaflow pipelines with 1,000+ concurrent computations (scaling toward 10k+) and per-solve failure isolation.
• Built end-to-end data lineage across forecast, strategy & trading — threading URN-style resource identity from ground data through ontology, data catalog, dbt, and Dagster to outcomes.

Built & scaled an AI product studio helping startups go from idea to production (0→1). Scaled engineering team to 35 organically, operated profitably, and led architecture across frontend, backend, cloud, data, and AI/ML systems.

• Built agentic AI systems using LangChain, RAG, tool-calling, and recommendation engines — deployed across client products with real-time inference and feedback loops.
• Delivered AI-driven platforms for clients across legal-tech (courtroom.ai), healthcare (dentalpracticeconnect.com), real-estate, manufacturing (koo.bz), electronics (sluicebox.ai) and more.
• Incubated HRHouz, an AI-driven ATS platform with agentic resume screening and candidate-matching flows. Won Antler's Shark Tank pitch event.
• Technical advising for startups — YearnMedia, HotMic, Liquidonate, Inductive Robotics, Smartpath, Courtroom, Forkast Restaurants.

Led a team of 5 open-source contributors, growing the engineering org to 20. Offered mentorship and technical guidance to the Open Source Project Steering Committee.

• Built multi-cloud DBaaS on Terraform, Kubernetes, and Helm — 8+ regions across AWS & GCP, 100+ customers.
• Stood up Prometheus/Loki/Grafana/PagerDuty observability stack; sustained 99.99% uptime over 6 months.
• Cut cloud spend via spot-instance migration and cluster right-sizing.
• Migrated CI/CD from TeamCity to GitHub Actions for both OSS and cloud pipelines.
• Led SOC 2 compliance end-to-end in under 3 months.

Led 5 platform engineers; promoted to Algorithms Platform Architect supporting 30+ engineers and managers on platform strategy.

• Architected the ML workflow and batch-execution platform (Futura scheduler + Flotilla executor) powering 15k+ daily workflows for 100+ data scientists.
• Saved ~$1.5M/quarter via bin-packing, autoscaling, and spot instances; improved cloud security posture from C+ to A+.
• Contributed to Docker-based microservice framework enabling rapid model deployment and configurable A/B testing.

Led engineering for a multi-cloud IaaS/PaaS abstraction layer for Webex — self-service CI/CD, observability, and datastore abstractions enabling 200+ developers.

• Built and operated Webex infrastructure across AWS, GCP, Rackspace, and OpenStack — 40+ regions, 3M+ DAU.
• Migrated legacy monitoring to cloud-native observability (ELK, Telegraf, InfluxDB, Grafana, Sensu, PagerDuty).
• Designed automated DR workflows; reduced failover time from 3 hours to 15 minutes.

Built a multi-tenant Identity Cloud Service (IDCS) on Oracle Public Cloud — SSO, LDAP, SAML, MFA, lifecycle management, and access gateways.

• Authored the Audit Manager microservice for identity event logging, compliance, and anomaly detection.
• Built Chef/Python orchestration for zero-downtime upgrades of 10+ microservices across global environments.
• Designed Docker-based integration testing deploying full IDCS stacks in isolated environments.

ML models and sensor-data pipelines for Snapdragon SoCs — thermal analysis, activity recognition, and gesture detection.

• Built thermal models used to diagnose and solve chip overheating on Snapdragon SoCs.
• Created automated classifier evaluation tools, cutting ML testing cycles from days to hours.
• Prototyped pickup-gesture detection for face unlock — real-time inference on accelerometer and gyroscope signals.

TA for Computer Architecture and Automata & Formal Languages.

Multi-screen IPTV apps (Android, iOS, Web) with Smooth Streaming, plus OCR/image-based UI test automation — deployed across 80+ global customer sites.

• Designed end-to-end IPTV backend and headend systems — satellite RF ingestion, stream transcoding, and API services.
• Cut testing from 2 weeks (5 testers) to 3 days (1 tester) through automated UI testing.

Developed 3D/4D medical imaging techniques for Thoracic Insufficiency Syndrome — modeling lungs and diaphragm from 2D MR images.

Programmed ABB Flexpicker with integrated vision systems for Cadbury factory packaging automation.

Built a computer-vision PCB quality checker — reduced inspection time from 7 min to 20 sec per board.